• Keine Ergebnisse gefunden

2 Theoretical Background

2.3 Salience-based approach

2.3.4 Psycholinguistic studies

Since 1970s psycholinguistic research on spoken languages has been investigating single and multiple factors affecting the salience of referents. Among those are the following: grammatical subject preference (Crawley & Stevenson 1990); first mention preference (Carreiras, Gernsbacher & Villa 1995; Gernsbacher & Hargreaves 1988;

Järvikivi et al. 2005); thematic role or semantic focusing of the verbs (Miltsakaki 2007;

Stevenson, Crawley & Kleinman 1994); implicit causality of the verbs (Hartshorne 2014;

Caramazza et al. 1977; Garvey, Caramazza & Yates 1974); type of connective words (Stevenson, Crawley & Kleinman 1994; Stevenson et al. 2000); information structural units such as topic and focus (Arnold 1998; Kaiser 2010; Colonna, Schimke & Hemforth 2012) as well as the interaction of multiple factors (i.e. semantic and syntactic) (Rose 2005).

Investigations of the factors influencing salience of the referents in the context of an anaphora resolution have only been done recently for sign languages by looking at the local utterances. Some of those studies focused on determining the influence of modality independent factors (i.e. first mention effect, verb causality) (Frederiksen & Mayberry 2017;

Frederiksen 2018; Wienholz et al. 2018b) while others examined how modality dependent factors (i.e. localization) affect comprehension of pronominal IX and other types of referential items (i.e. bare nouns) (Wienholz et al. 2018a; Wienholz et al. 2018c).The results of these studies which are relevant for the current work are presented below.

56

Studies on spoken languages revealed that the subject arguments occupy more prominent positions that other sentential arguments, therefore reduced referential expressions are usually observed to refer to the subject of a sentence (Givón 1984; Gordon, Grosz & Gillom 1993; Lambrecht 2000). Other studies showed that it is not only the subjecthood but indeed the initial order of mention which increases the prominence of referential expressions (Gernsbacher & Hargreaves 1988; Carreiras, Gernsbacher & Villa 1995). The effect of the order of mention and/or subjecthood on the production and comprehension of referential expressions was tested for two sign languages (i.e. ASL and DGS), and revealed differing results.

Local contexts were examined in a pilot study on ASL via sentence continuation task focusing on production and comprehension of pronominal IX (Frederiksen & Mayberry 2017). Given the relevance of the local contexts and the design for the current study (Chapter 5), some details of this task are provided already here to set the ground for the further discussion. Frederiksen & Mayberry (2017) investigated whether referentially anchored and unanchored pronominal IX follow structural constraints (i.e. subject preference) in identification of their referents. Therefore, they conducted two tasks: (i) free, and (ii) controlled sentence continuation. The first task contained 96 prompt sentences presented in two conditions either localization or neutral localization (i.e. no localization or localization in the neutral area of the signing space in front of the torso of the signer) of sentential arguments. The second task contained 16 prompt sentences with neutral localization of the arguments followed by sentences starting with referentially unanchored pronominal IX.Four signers of ASL were instructed to watch the videotaped sentences and continue them without any restrictions in the free continuation task while in the controlled setting they were supposed to continue sentences already starting with pronominal IX.The findings for free continuations indicate that when pronominal IX was chosen to continue the prompt sentences

57

in localization condition it was co-referential with the subjects in slightly more cases (35,27%) when compared to the objects (30,72%)34. However, in neutral condition IX was identified to refer to objects more frequently (45,46%) than to subjects (22,55%). In controlled continuations, signers had a slight preference to identify and continue IX initial sentences with the objects (52,12%) than with the subjects (47,88%). Overall, the findings suggest that referentially unanchored pronominal IX signs are more frequently identified to refer to objects than to subjects.

We conducted an event-related potential study examining the presence of a first mention effect during pronoun resolution in DGS in ambiguous contexts (Wienholz et al.

2018b). Therefore, participants were presented with sentence sets containing two referents without overt localization in the first sentences and pronominal IX at the beginning of the second sentence directed to either the right (ipsilateral) or left (contralateral) side of the signing space. Results show an N400 in the contralateral compared to the ipsilateral condition suggesting increased processing costs for the contralateral IX sign, which refers to the second-mentioned referent. Thus, it was interpreted as supporting evidence for a first mention effect in DGS. Given that DGS signers were shown to follow a right-left default pattern while assigning referents such that the first referent is overtly or covertly associated with the ipsilateral and the second referent with the contralateral area in the signing space (Steinbach & Onea 2016; Wienholz et al. 2018a). Hence, directing a pronoun to one of those sides would identify its referent either as the first or second-mentioned one.

In yet another study, we examined whether overt manual localization increases the prominence and hence the accessibility of a discourse referent and how this interacts with its grammatical role in DGS (Wienholz et al. 2018c). Using eye tracking and a modified

34 Note that in the free continuation task referential expressions other than pronominal IX (e.g. bare nouns) were produced as well. Those findings will be discussed further in Chapter 5, Section 5.4.

58

version of the Visual World Paradigm, participants were presented with two pictures representing the discourse referents contained in the stimulus sentence appearing in a video.

Each video included short discourses that introduced two discourse referents with varying their overt localization in a first sentence while a subsequent second sentence started with one of the referents, i.e. as a bare noun. Analyzing proportions of target looking using mixed-effects models revealed increased looks to the target referent for conditions containing overt localization of both referents or only localizing the subject. This suggests that overtly localizing a referent indeed enhances its accessibility, but only if the referent occurs in the subject position. Moreover, localization seems to accentuate a referent in a similar way as prosodic focus in spoken languages. Thus, the combined factors of localization and subject preference lead to facilitatory processing of referential expressions co-referential with the focused referent.

To wrap up, the sign language studies that have investigated the factors that influence salience in the production and the comprehension of referential expressions were mainly conducted on well investigated sign languages such as ASL, and were primarily focused on the production of these expressions in different genres of discourse. As mentioned earlier, the effect of one particular convention/factor (i.e. accessibility) on the salience of referential expressions was investigated. In production, the highly accessible referents were observed to be referred to via pronominal IX with varying number depending on the type of a genre and the complexity of the discourse. In comprehension, object preference for ASL while subject/first mention preference for DGS were identified to be the factors influencing interpretation of pronominal IX in controlled local contexts (for an overview, see Table 2.4 below).

59

Table 2.4: An overview of the factors determining salience of referential expressions proposed in sign language literature

60